Small raised pimples on the outer surface, sometimes glossy, sometimes with a tiny tail. They catch the light and they catch a fingernail, and on a part that is otherwise good they are the only thing anyone notices.
Where are yours?
This matters more than what they look like, because different restarts produce them.
All in one vertical line up the part. That is the layer start point, and it is a different problem with different answers — see visible Z seam.
Scattered around the wall with no pattern. Each one marks a travel move: the extruder stopped, the head went somewhere, and flow restarted at the far end. This page is about those.
Only where the wall meets a hole or a cut-out. Same cause, made obvious by geometry — internal perimeters mean many more starts and stops per layer.
On the top surface rather than the sides. Look at rough top surface instead; that is a flow calibration issue rather than a restart one.
Why a restart leaves a lump
Molten plastic in a nozzle is not an incompressible fluid connected rigidly to the drive gear. There is compliance in the melt and in the filament column, so pressure builds up over the first millimetres of a move and decays over the last few.
When the extruder stops, the pressure left in the melt keeps pushing plastic out for a moment. When it starts again, there is a lag before pressure rebuilds. If the firmware compensates for neither, you get an excess at the end of one path and a starved beginning to the next — and the excess is the bump you can feel.
Some slicers add an explicit "extra restart distance" on top of retraction, which pushes a little more filament in when printing resumes. If it is set above zero on a machine that does not need it, it manufactures this defect directly.
The fix that actually addresses the cause
Calibrate pressure advance — called linear advance in some firmware, and the same idea in both. It tells the controller to lead the extruder ahead of the motion planner while accelerating and behind it while decelerating, so the melt pressure tracks the speed instead of lagging it.
A calibration print produces a series of lines at increasing values; you pick the one where the ends and the middle look alike. The value is dimensionless and firmware-specific, and the geometry matters: a direct drive machine's number usually lands in the low hundredths, while a bowden setup, with far more compliance in its long filament column, commonly needs something like an order of magnitude more.
This one calibration also improves corner bulging and the seam, because all three are the same physics observed at different moments.
The rest, in order
- Set extra restart distance to zero before adjusting anything else. It is the most common cause of self-inflicted zits.
- Enable a small wipe at the end of each loop, so the nozzle drags the surplus back along the path it just printed rather than depositing it.
- Reduce retraction distance if it is unnecessarily long. Every millimetre pulled back is a millimetre that has to be re-pressurised.
- Turn on coasting only if the above are exhausted. It stops extruding slightly before the end of a path and relies on residual pressure to finish it, which trades a blob for a possible gap — useful, but a compensation rather than a correction.
A test that tells you whether you fixed it
Print a small cube with one bolt hole through a side face. The hole forces two extra starts and stops per layer on the wall around it, so bumps concentrate there and become countable. Photograph that face in raking light before and after a change, and compare the photographs rather than your memory of the part. Restart defects are exactly the kind of small, distributed flaw that people convince themselves has improved when it has not.
What will not help
Raising the temperature to "smooth them out". A hotter, thinner melt oozes more freely at every stop, so this reliably makes it worse, and it brings stringing along with it.
Sanding is the other trap. It works, and on a part with many of these it is an afternoon that a ten-minute calibration would have saved.
The materials that show it
PETG produces the most visible bumps because its melt is viscous and stringy and it holds pressure well. Glossy PLA shows small ones clearly for optical reasons rather than physical ones. Matte and fibre-filled grades hide them almost entirely — worth knowing when choosing a filament for a display piece where the geometry has a lot of small features and therefore a lot of restarts.